tool_call_malformedTier 1 · 70% confidence

ai-agents-tool-call-malformed-when-using-glm-5-or-glm-4-7-models-with-mtp-multi--15e1c737

agent: ai_agents

When does this happen?

IF When using GLM-5 or GLM-4.7 models with MTP (Multi-Token Prediction) spec decoding enabled in vLLM, tool call JSON output is malformed (e.g., missing closing braces) leading to parse failures.

How others solved it

THEN Disable MTP spec decoding for GLM models when using tool calling, or downgrade to vLLM 0.15.0 where the issue is absent. In vLLM, set the `--enable-mtp-decoding` flag to False or remove it from the server configuration.

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics